COMPARATIVE ANALYSIS OF STROKE CLASSIFICATION USING THE K-NEAREST NEIGHBOR DECISION TREE, AND MULTILAYER PERCEPTRON METHODS

Abstract

Stroke has become a serious health problem; the main cause of stroke is usually a blood clot in the arteries that supply blood to the brain. Strokes can also be caused by bleeding when blood vessels burst and blood leaks into the brain. In one year, about 12.2 million people will have their first stroke, and 6.5 million people will die from a stroke. More than 110 million people worldwide have had a stroke. Handling that is done quickly can minimize the level of brain damage and the potential adverse effects. Therefore, it is very important to predict whether a patient has the potential to experience a stroke. The K-Nearest Neighbor, Decision Tree, and Multilayer Perceptron algorithms are applied as a classification method to identify symptoms in patients and achieve an optimal accuracy level. The results of making the three algorithms are quite good, where K-Nearest Neighbor (K-NN) has an accuracy value of 93.84%, Decision Tree is 93.97%, and Multilayer Perceptron (MLP) is 93.91%. The best accuracy value is the Decision Tree algorithm with an accuracy difference of no more than 0.10% with the two algorithms used.